Research On Millimeter Wave Wi-Fi Beam Management Based On Machine Learning

With the increasing demand for ultra-high-speed wireless communication,the microwave frequency band cannot meet future transmission needs due to limited spectral resource.The60 GHz wireless local area network(WLAN)with 7 GHz unlicensed spectrum is considered as a viable candidate to support gigabit transmission speed.Millimeter wave(mm Wave)communication is becoming one of the most prominent technologies in both the fifthgeneration mobile communication system and the next generation WLAN(e.g.,IEEE802.11ad/ay).However,mm Wave band faces many challenges,such as severe pathloss,high oxygen absorption,and blockage.Therefore,mm Wave communication needs beamforming technology by multi-antenna arrays to maintain link budget,which is critical to overcome the attenuation challenge in future mm Wave communication systems.The beamforming can concentrate the transmit power and receive region on a narrow beam,which makes the mm Wave has the advantages of strong directivity and high spatial resolution,so it becomes the core technology of mm Wave communication.In this thesis,the beam training and beam tracking mechanisms in IEEE 802.11 ad standard were studied,and an effective and intelligent algorithm is designed to perform the beam management and mobility management in mm Wave Wi-Fi.In this thesis,the channel access,beam training and beam tracking mechanisms of802.11 ad are studied in depth.Aiming at the problem of over-consuming bandwidth resource and high time overhead in the beam sweep process during BTI phase,proposes a scheme based on high-low-frequency cooperative framework.By sending beacon frame on the lowfrequency channel,and then sending smaller sweep frame on the high-frequency channel,high-frequency occupation time and beam sweep overhead are reduced.Aiming at the problem of high probability of collisions in dense user scenarios during A-BFT phase,resulting in long access delay and inefficient training performance,proposes two solutions.One is to increase the number of accessible time slots by extending the A-BFT time slot,and to classify STAs for access attempts.The other is to use reinforcement learning algorithm to train STA to actively adjust its probability distribution to select random numbers according to the network environment,so that the probability of the same random number being repeatedly selected is reduced.Compared with the legacy 802.11 ad A-BFT,simulation results show that two solutions both can greatly reduce the collision probability in dense user scenarios and significantly improve beamforming training efficiency.Furthermore,the beam tracking process is also of great concern in mm Wave system which can significantly increase the user's received signal power in high-speed communications.However,the existing algorithm has the problem of long training time,which will seriously affect the beam tracking performance of mobile users.Taking into account the attribute of user's movement behavior,this thesis proposes a new mm Wave beam tracking method based on deep learning network.The LSTM network uses historical trajectory information to predict the user's moving position at the next moment to determine the target beam in the beam tracking process and measure whether the predicted beam is available.Finally,the beam tracking scheme in 802.11 ad and the scheme proposed in this thesis are compared and analyzed by using MATLAB simulation software.The simulation results show that the number of beams traversed by the scheme proposed in this thesis is reduced,and the link robustness is improved.